PHF2 histone demethylase prevents DNA damage and genome instability by controlling cell cycle progression of neural progenitors

Stella Pappaa, Natalia Padillab,c,1, Simona Iacobuccia,1, Marta Viciosoa, Elena Álvarez de la Campab,c, Claudia Navarroa, Elia Marcosa, Xavier de la Cruzb,c,2, and Marian A. Martínez-Balbása,2

aDepartment of Molecular Genomics, Instituto de Biología Molecular de Barcelona, Consejo Superior de Investigaciones Científicas, 08028 Barcelona, Spain; bClinical and Translational Bioinformatics, Vall d’Hebron Institute of Research, E-08035 Barcelona, Spain; and cInstitut Català per la Recerca i Estudis Avançats, 08018 Barcelona, Spain

Edited by Robert E. Kingston, Massachusetts General Hospital/Harvard Medical School, Boston, MA, and approved August 6, 2019 (received for review February 25, 2019) Histone H3 lysine 9 methylation (H3K9me) is essential for cellular terestingly, PHF2 mutations have been found in patients with homeostasis; however, its contribution to development is not well autism spectrum disorder (ASD) (22–24). Despite PHF2 high established. Here, we demonstrate that the H3K9me2 demethy- expression in the neural tube and its implication on mental dis- lase PHF2 is essential for neural progenitor proliferation in vitro eases, PHF2 involvement in neural development has not yet and for early neurogenesis in the chicken spinal cord. Using genome- been explored. wide analyses and biochemical assays we show that PHF2 controls In this study, we analyzed the role of PHF2 in neural stem cell the expression of critical cell cycle progression , particularly (NSC) biology. We found that PHF2 is essential for progenitor those related to DNA replication, by keeping low levels of H3K9me3 proliferation in vitro and in vivo, in the chicken spinal cord. at promoters. Accordingly, PHF2 depletion induces R-loop accu- PHF2 binds and regulates cell cycle promoters, particularly mulation that leads to extensive DNA damage and cell cycle arrest. These data reveal a role of PHF2 as a guarantor of genome stabil- those involved in DNA replication and cell cycle progression. ity that allows proper expansion of neural progenitors during Moreover, PHF2 depletion induces R-loop accumulation, DNA development. damage, and cell cycle arrest. These data reveal a role for PHF2 as a safeguard of genome stability during development. histone demethylation | transcription | neuronal progenitor Results proliferation | PHF2 PHF2 Binds Promoters and Mediates H3K9me2 Demethylation. To uring neural development, multipotent progenitor cells self- gain insight into the biological substrate of PHF2 in NSCs, we Drenew and ultimately originate specialized neurons and glial performed chromatin immunoprecipitation coupled with sequencing cells (1, 2). The chromatin acting factors are essential players in both proliferation and cell differentiation events during embryo Significance development. This epigenetic regulation is achieved by stabiliz- ing chromatin structure that allows establishing heritable gene During early neurogenesis, progenitors maintain the balance expression patterns. The epigenetic control is mainly mediated between self-renewal and differentiation. This equilibrium is by covalent modifications of histones and DNA (3). Recently, preserved by the coordinated action of developmental cues histone methylation/demethylation has received special attention and epigenetic mechanisms. We analyzed the role of the as an essential regulator of gene expression and genome stability H3K9me2 histone demethylase PHF2 in neural stem cells bi- during development (4). One critical histone modification during ology in vitro and in early neurogenesis in the chicken neural development is dimethylation of histone H3 at lysine 9 (H3K9me2), tube. We found that PHF2 is essential for proliferation of pro- implicated in the silencing of genes playing important roles in genitors. Additionally, PHF2 maintains the genome integrity, chromatin homeostasis and development (5). In embryonic preventing R-loop accumulation and DNA damage. Our work stem (ES) cells, it has been proposed that H3K9me2 increases contributes to the clarification of both the physiological and across the genome as cells differentiate and acquire lineage molecular roles of PHF2 during early neurogenesis. It opens specificity (6), although this is contentious (7). In addition, avenues to understand the role of H3K9 methylation not only H3K9me2 together with H3K9me3 are essential components of in development but also in diseases such as autism, cancer, and the constitutive heterochromatin (8–10). obesity where PHF2 has been involved. Despite the importance of this modification during develop- ment, little is known about the role of the enzymes responsible Author contributions: S.P., X.d.l.C., and M.A.M.-B. designed research; S.P., S.I., M.V., C.N., and E.M. performed research; X.d.l.C. and M.A.M.-B. contributed new reagents/analytic for this mark in neurogenesis. PHF2 is a member of the KDM7 tools; S.P., N.P., S.I., M.V., E.Á.d.l.C., C.N., E.M., X.d.l.C., and M.A.M.-B. analyzed data; and histone demethylase (HDM) family (11). It contains a plant home- M.A.M.-B. wrote the paper. odomain (PHD) in the N-terminal and the Jumonji-C (JMJC) do- The authors declare no conflict of interest. main, which has demethylase enzymatic activity (12). Biochemical This article is a PNAS Direct Submission. studies demonstrated that PHF2 demethylates H3K9me2 upon Published under the PNAS license. interaction with H3K4me2/3 through its PHD domain (13). Data deposition: RNA-seq and ChIP-seq data have been deposited in the GEO database PHF2 was first identified as a candidate gene for hereditary sen- under accessions GSE122264 and GSE122263, respectively. sory neuropathy type I (14) because it is expressed at high levels in 1N.P. and S.I. contributed equally to this work. the neural tube and dorsal root ganglia (15). The physiological 2To whom correspondence may be addressed. Email: [email protected] or mmbbmc@ role of PHF2 in vivo is not yet clear, but it is a coactivator of ibmb.csic.es. multiple transcription factors (11). Working with them, PHF2 This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. regulates various differentiation processes (16–18). Alterations in 1073/pnas.1903188116/-/DCSupplemental. PHF2 have been identified in several cancer types (19–21). In- First published September 5, 2019.

19464–19473 | PNAS | September 24, 2019 | vol. 116 | no. 39 www.pnas.org/cgi/doi/10.1073/pnas.1903188116 Downloaded by guest on September 29, 2021 (ChIP-seq) experiment. We detected 5,992 peaks normalized to research tool. We identified as the top 3: Homez, YY2, and the input (P value of 1e-10) in ChIP data for PHF2. SI Appendix, E2F4 DNA binding motifs (Fig. 1E). These transcription factors, Fig. S1A shows validation of ChIP-seq results by qPCR for a ran- particularly E2F4 and YY2, are essential to control progenitor dom set of PHF2 targets. Then, we examined the genomic distri- self-renewal and cell cycle progression (26–28), suggesting a po- bution of the PHF2 peaks. Our results showed that 97% of PHF2 tential role of PHF2 in cell proliferation. To reinforce these data, peaks localized on gene promoters, around the transcription start we analyzed the colocalization of PHF2 and E2F4 using pre- site (TSS) (Fig. 1A), as it has been reported in a different cellular viously published ChIP-seq data in HeLa S3 cells. The results context (25). showed that the 50.3% (P value <2.2e-16) of E2F4-bound re- It has been proposed that PHF2 through its PHD domain gions, respectively, also contained PHF2 (Fig. 1F). These results interacts with H3K4me2/3 histone marks; thus, we analyzed the reveal a potential role of PHF2 in regulating E2F-mediated colocalization of PHF2 and H3K4me2/3 at the genome-wide transcription as it has been described for another KDM7 family level using previously published H3K4me2/3 ChIP-seq data from member, PHF8 (29). Accordingly, (GO) analysis of NSCs. Doing that, we identified 5,978 (99.7%) (P value <2.2e-16) PHF2-bound regions indicated that PHF2 was associated with the and 5,983 (99.8%) (P value <2.2e-16) PHF2-bound regions promoter region of genes involved in cell cycle regulation, DNA that also contained H3K4me3 or H3K4me2, respectively (Figs. 1 repair, RNA processing, and chromatin organization, among B–D and 2B and SI Appendix,Fig.S1B). The high overlapping others (Fig. 1G). suggests a required H3K4me2/3 mark for effective PHF2 binding Previous studies demonstrated that PHF2 demethylates mainly at promoters as it is described (13, 25). To better understand H3K9me2 (13, 25). We thus analyzed the distribution of H3K9me2 how PHF2 is targeted to the chromatin, we performed bio- in NSCs by performing ChIP-seq experiments and compared it informatics analysis that revealed the most statistically significant with PHF2 genomic-associated regions. As expected for a H3K9me2 predicted PHF2 binding sites using the Homer de novo motif demethylase, the vast majority (99.99%) (P value <2.3e-6) of BIOCHEMISTRY

Fig. 1. PHF2 binds promoters in NSCs. (A) Genomic distribution of PHF2 ChIP-seq peaks in NSCs showing that PHF2 mainly binds promoter regions around the TSS. (B) Heatmaps depicting PHF2 binding to H3K4me3- and H3K4me2-marked promoters in NSCs 3 kb around the TSS. Scales indicate ChIP-seq intensities. (C)Venn diagrams showing overlap between PHF2-bound and H3K4me3 and H3K4me2-marked regions. P value is the result of an equal proportions test performed between H3K4me3 and H3K4me2 peaks and a random set. (D) IGV capture showing PHF2, H3K4me3, and H3K4me2 peaks in E2f3 gene in NSCs. (E) Motif enrichment analysis of PHF2 ChIP-seq peaks in NSCs using the Homer de novo motif research tool showing the 3 top enriched motifs. (F) Heatmap representation of PHF2 and E2F4 co- occupied regions (Left) and Venn diagrams (Right) showing peak overlapping between PHF2-bound and E2F4-bound regions in published ChIP-seq in HeLa S3 cells. (G) Gene ontology analysis showing the biological process of the PHF2-bound genes was performed using as a background the whole Mus musculus genome.

Pappa et al. PNAS | September 24, 2019 | vol. 116 | no. 39 | 19465 Downloaded by guest on September 29, 2021 Fig. 2. PHF2 mediates H3K9me2 demethylation. (A) Venn diagram showing overlap between PHF2-bound and H3K9me2-marked regions in NSCs. (B) Clustered heatmap showing Pearson correlation of H3K9me2, H3K4me2, H3K4me3, and PHF2 ChIP-seq samples based on read coverage within genomic regions. (C) IGV capture showing PHF2 peaks and H3K9me2-marked regions in the Mcm6 gene. The promoter region in the box shows no H3K9me2 enrichment where PHF2 binds. (D) Venn diagram depicting overlap between PHF2-bound and H3K9me3-marked regions in NSCs. (E) NSCs were infected with lentivirus expressing shRNA control (shControl) or shRNA specific for PHF2 (shPHF2). After 48 h, total extracts were prepared and the PHF2 and TUBULIN levels were determined by immunoblot. (F) Immunostaining of shControl and shPHF2 cell lines. Cells were fixed and stained with PHF2, H3K9me2, and H3K9me3 antibodies and DAPI (Scale bar: 20 μm.) Zoom in showing H3K9me3 foci in shControl and shPHF2. Cell fluorescence for H3K9me2 staining and H3K9me3 foci was measured using ImageJ. Boxplots represent the quantification of the fluorescence intensities as well as the number of H3K9me3 foci/cell in shControl and shPHF2 cells. **P < 0.01; ***P < 0.001 (Student’s t test). (G) ChIP of H3K9me3 in shControl and shPHF2 cells was analyzed by qPCR at the indicated gene promoters that were identified by PHF2 ChIP-seq as PHF2 targets. An intragenic region of the Gda gene marked by H3K9me2/3 but devoid of PHF2 was used as negative control. Data from qPCR were normalized to the input and expressed as fold enrichment over the data obtained in shControl. Results are the mean of 2 to 3 biological independent experiments. Error bars represent SEM. *P < 0.05; **P < 0.01 (Student’s t test).

PHF2-bound regions were totally excluded from the H3K9me2 the expression of its homologous PHF8 (SI Appendix, Fig. S2B). positive regions in NSCs (Fig. 2 A–C and SI Appendix,Fig.S2A). After viral transduction, the levels of H3K9me2 were analyzed by Accordingly, clustered heatmaps showed a clear exclusion of the immunofluorescence assays. In Fig. 2F we observed a slight in- H3K9me2 mark and H3K4me2/3-associated regions (Fig. 2B). crease in H3K9me2 levels as it has been demonstrated in other Moreover, a strong overlapping of PHF2 peaks with H3K4me2/3- cellular contexts (13). As the H3K9me2 mark serves as a substrate associated regions was observed, in agreement with the data pre- for the histone methyltransferase SUV39H, we tested the levels of sented in Fig. 1 B and C. Altogether these results suggest that H3K9me3 in PHF2-depleted NSCs. Results in Fig. 2F clearly PHF2 binds to genomic regions marked with histone modifications showed an increase in H3K9me3. Intriguingly, both the intensity related to transcriptional activation. Similarly to H3K9me2, H3K9me3- and the number of H3K9me3 foci increased upon PHF2 depletion. enriched regions identified by previously published ChIP-seq were Accordingly, a clear accumulation of HP1α was also detected in excluded from the PHF2 positive regions in NSCs (Fig. 2D). To PHF2-depleted cells (SI Appendix,Fig.S2C) without increase in evaluate the impact of PHF2 depletion on histone modifications other histone marks such as H3K4me3 (SI Appendix,Fig.S2D). in vivo, NSCs were transduced with lentivirus containing specific These data strongly suggest that depletion of PHF2 led to an PHF2 shRNA that efficiently decreased the PHF2 levels (Fig. 2E increase in heterochromatin-related marks at global levels. Thus, we and SI Appendix, Fig. S2B) or with a control shRNA (shControl) sought to test whether PHF2 is also important to prevent H3K9me3 (see Materials and Methods). The reduction of PHF2 did not affect increase at a local level. To do so, we chose some PHF2-target regions

19466 | www.pnas.org/cgi/doi/10.1073/pnas.1903188116 Pappa et al. Downloaded by guest on September 29, 2021 identified in the ChIP-seq experiment and tested the effect of PHF2 PHF2-associated transcriptional profile by RNA-seq. We found depletion on H3K9me3 levels by ChIP-qPCR assays. A clear 1,729 transcripts that significantly changed their expression increase in the H3K9me3 mark was observed (Fig. 2G), at the an- (log2 fold change [FC] > 0.8 and FC < −0.8 and P value <0.01), alyzed promoters, without affecting the intragenic region of the Gda in the 2 biological independent experiments in PHF2-depleted gene (a non-PHF2 target used as a negative control), in accordance NSCs (shPHF2) compared with control (shControl) cells (Fig. 3 with the global increase noticed in Fig. 2F. These data strongly A and B). Among these, 791 (45.8%) were down-regulated and suggest that PHF2 prevents H3K9me3 accumulation, limiting ec- 938 (54.2%) up-regulated upon PHF2 depletion (Fig. 3C). These topic heterochromatin formation, particularly at promoter regions. results were confirmed by depletion of PHF2 by using another shRNA against a distinct region of PHF2 (see Materials and PHF2 Regulates Cell Cycle Gene Transcription. To gain further Methods) as we tested by qPCR analysis of some randomly knowledge into PHF2’s function in NSCs, we identified the chosen genes (SI Appendix, Fig. S3A). Next, we identified the BIOCHEMISTRY

Fig. 3. PHF2 regulates cell cycle gene transcription. (A) Volcano plot represents PHF2 transcriptional targets identified by RNA-seq in shControl and shPHF2 NSCs. The green dots represent all of the genes with false discovery rate (FDR) < 0.01 and log2 FC > 0.8 and log2 FC < −0.8. (B) Heatmap showing the top 20 regulated genes identified by RNA-seq in shControl and shPHF2. Two biological replicates of shPHF2 cells were used for RNA-seq. All of the genes showed P value <0.01 and log2 FC > 0.8 and log2 FC < −0.8. (C) Diagram showing the number and percentage of up- and down-regulated genes in the RNA-seq experiment comparing shControl and shPHF2 NSCs. (D) Venn diagram showing overlapping between PHF2-bound genes and PHF2 transcriptional targets. A total of 34.8% of differentially expressed genes identified by RNA-seq with log2 FC > 0.8 and log2 FC < −0.8 were also PHF2 direct targets identified by ChIP-seq. From these, 34% were up-regulated and 66% down- regulated. (E) Graph depicting the percentage and number of up-regulated and down-regulated genes in the shPHF2 according to the RNA-seq that contain PHF2 in their promoter classified by log2 fold change. (F) Gene ontology analysis showing the “biological process” ofthePHF2directtargetswas performed using as a background the whole M. musculus genome. (G) IGV capture showing PHF2 peaks and RNA levels in shControl and shPHF2 in the Orc1 gene.

Pappa et al. PNAS | September 24, 2019 | vol. 116 | no. 39 | 19467 Downloaded by guest on September 29, 2021 PHF2 direct transcriptional targets by comparing the PHF2- PHF2 Regulates Cell Proliferation. To further analyze the potential associated transcriptional profile with the ChIP-seq data. role of PHF2 in cell proliferation, we examined the consequences Among the genes that showed a PHF2 dependency for tran- of its reduction. PHF2-depleted NSCs exhibited a striking de- scriptional regulation (log2 FC > 0.8 and log2 FC < −0.8) in the crease in cell growth (Fig. 4A). Moreover, flow-cytometry analysis RNA-seq experiment, 601 (34.8%) were bound by PHF2 (Fig. demonstrated a delay in G1/S transition (%G1 shControl 42.2%, 3D). Interestingly, the proportion of direct down-regulated shPHF2 53.5%) upon PHF2 depletion (Fig. 4B). Moreover, the transcripts (397, 66%) was higher than the direct up-regulated levels of the Mcm2 factor (the putative helicase essential for DNA ones (204, 34%) (Fig. 3D). This percentage was almost main- replication initiation and elongation in eukaryotic cells) phos- tained at different FCs (Fig. 3E). Moreover, half of the down- phorylated at S40 required for the initiation of DNA replication regulated genes in the RNA-seq experiment (50.2%) contain were lower in PHF2-depleted cells compared with control pro- C PHF2 bound at the promoter region as we determined in the genitors (Fig. 4 ). Interestingly, the expression of PHF2 was SI Appendix B regulated throughout the cell cycle with higher levels at G1/S ChIP-seq experiment ( , Fig. S3 ), while only 21.7% SI Appendix A of the up-regulated genes were direct targets of PHF2 (SI Ap- transition ( ,Fig.S5) and its recruitment to the pendix, Fig. S3C). These data are in accordance with the acti- S-phase-activated gene promoter ORC1 also increased at G1/S transition (SI Appendix, Fig. S5B). These data reinforce the idea vator role proposed for PHF2 (17). Nevertheless, the fact that that PHF2 contributes to cell cycle progression by facilitating the 34% of the direct targets became up-regulated upon PHF2 de- expression of cell cycle progression genes, particularly those in- pletion, suggest a potential role of PHF2 contributing to tran- volved in DNA replication. scriptional repression or preventing activation as it has been demonstrated for another member of the KDM7 family, PHF8, PHF2 Depletion Blocks Neurogenesis in the Spinal Cord. The findings in a different cellular context (30). Enrichment analysis of GO described above support the idea that PHF2 activates genes es- terms over the 601 PHF2 direct target genes showed that the sential for neural progenitor proliferation (Fig. 3F and SI Ap- most enriched were associated with cell cycle categories, par- pendix, Fig. S4 A and B). Thus, we sought to test this notion in ticularly G1/S transition (E2f2/3/7/8, Cdc7, Cdc25a, Cdk4, and vivo using the chicken embryo neural tube as a model. Structur- Mcm3/4/8), DNA replication (Orc1/2/6 and Pcna), mitosis (Cdk1, ally, 2 main zones can be distinguished in a transversal section of Smc2/3/4, Aurkb, and Topo2a), as well as chromatin activity the neural tube: the ventricular zone (VZ), where proliferating (Cenpa, Kdm1b, Hat1, Parp1, and Prmt5) (Fig. 3 F and G and SI progenitors reside, and the mantle zone (MZ), where final dif- Appendix, Fig. S4A). Intriguingly, some of them were E2F target ferentiated neurons accumulate (Fig. 4D). To analyze the function genes (Ccnd1, Cdc25a, Pcna, Mcm3/4/6/8, and Smc4, including of PHF2 in early neurogenesis, we first cloned 2 shRNAs for chick E2f family genes) (Fig. 3 F and G and SI Appendix, Fig. S4B). To PHF2 (cPHF2) in a bicistronic vector containing GFP sequence, further characterize the differences between control and PHF2 which reduced PHF2 levels (SI Appendix,Fig.S5C). Then, we knockdown (KD) NSCs, we performed a GO enrichment anal- knocked down PHF2’s expression in Hamburger and Hamilton ysis of down-regulated, up-regulated, and unregulated direct stage 10–12 (HH10–12) chicken embryo spinal cords by in ovo target genes to identify those biological processes most sensitive electroporation of the shRNAs for cPHF2 or a control shRNA to PHF2 depletion. The results revealed that PHF2 down- (shControl). Remarkably, cPHF2 KD resulted in a neural tube regulated genes were strongly associated with cell cycle pro- reduced in size (Fig. 4E and SI Appendix,Fig.S5D). As the gression, chromatin activity, and DNA repair among others (SI 2 cPHF2 shRNAs gave the same phenotype, we chose the Appendix, Figs. S3D and S4B). The same analysis of the up- shcPHF2, that provided better cPHF2 depletion (SI Appendix,Fig. regulated genes did show functional categories related to mor- S5C), to perform the rest of the experiments. The observed re- phogenesis, signal transduction, and developmental process (SI duced size of the neural tube could mainly be due to cell death or Appendix, Fig. S3E) or to RNA processing and chromatin or- to proliferation defects. As we did not observe apoptotic cells in ganization among others in the case of unregulated ones (SI the electroporated neural tubes and our previous results indicated Appendix,Fig.S3F). It is important to notice that the number of that PHF2 is essential for neural progenitor proliferation in vitro, we hypothesized that cPHF2 plays an active role in maintaining genes and the statistically significant values of up-regulated and – unregulated gene-associated categories were much lower (P val- neuroblast proliferating. To explore this idea, HH10 12 neural ues ranging from −log10 2–6) compared with the down-regulated tubes were electroporated with shcPHF2 or shControl and the effect on neural progenitor proliferation was analyzed after 48 h. genes (P values −log10 4–23). This fact, led us to focus on the We evaluated neural progenitor entry into mitosis by analyzing the potential role of PHF2 as a transcriptional activator of cell cycle presence of H3S10p. Neural tubes electroporated with shcPHF2 and DNA repair genes in this particular neural context. Finally, showed a reduction in H3S10p (PH3)-positive mitotic cells (ratio we confirmed by immunoblot the decrease of some of PH3+ cells electroporated [EP] side/control side: shControl whose genes were down-regulated in PHF2-depleted cells (SI ± ± P < F Appendix C 106.5 36.7, shcPHF2 65.3 21; 0.05) (Fig. 4 ). Accordingly, ,Fig.S4 ). Interestingly, we did not detect any reduction reduction of the neural progenitor marker SOX2 (ratio of SOX2+ in the histone protein levels, although their mRNAs were down- ± ± SI Appendix D cells EP site/control site: shControl 101.6 10.3, shcPHF2 67.3 regulated upon PHF2 depletion ( ,Fig.S4 ), probably 14; P < 0.05) (Fig. 4F) indicates that cPHF2 is required for neural due to the tight control of the histone levels inside the cells. Al- progenitor self-renewal in the neural tube. We next explored the together, these data strongly suggest that PHF2 binds to the cell possibility that the inhibition of proliferation observed upon cycle gene promoters to fine tune their chromatin activity and cPHF2 reduction would also correspond with a premature dif- facilitate their transcription. ferentiation of neuroblasts. Neural tubes electroporated in ovo We next sought to analyze the transcriptional consequences of with shcPHF2 and stained for TUJ1, a neuronal differentiation PHF2 overexpression in NSCs. To do that, we established a NSC marker, showed neither premature differentiation nor ectopic line that expressed PHF2 wild type (WT) upon addition of localization (Fig. 4F). Only a clear reduction of TUJ1-positive doxycycline (see Materials and Methods). We induced the PHF2 cells was observed, probably reflecting the decrease on the pro- expression and the levels of some direct PHF2 targets, identified genitor population. Finally, we investigated whether the neuro- in the RNA-seq and ChIP-seq experiments, were analyzed by blast proliferation impairment affected similarly to all neural qPCR. No changes or slight alterations of the expression (up and subpopulations along the dorsal–ventral axis. To do that, we down) were observed upon PHF2 overexpression (SI Appendix, depleted cPHF2 and analyzed different populations by using Fig. S4E). different neural markers: MNR2 for ventral neurons (motoneurons)

19468 | www.pnas.org/cgi/doi/10.1073/pnas.1903188116 Pappa et al. Downloaded by guest on September 29, 2021 BIOCHEMISTRY

Fig. 4. PHF2 depletion blocks neurogenesis in the spinal cord. (A) Growth curve showing the proliferation rate of NSCs infected with lentivirus expressing shRNA control (shControl) or shRNA for PHF2 (shPHF2) from 0 to 72 h. (B) Flow cytometry analysis of NSCs control and PHF2-depleted cells previously stained with propidium iodide. (C) Total protein extracts were prepared from NSCs infected with lentivirus expressing shRNA control (shControl) or shRNA specific for PHF2 (shPHF2) and the phospho-MCM2 (p-MCM2) and TUBULIN levels were determined by immunoblot. (D) Schematic representation indicating the regions occupied by proliferating neural progenitors (VZ) and postmitotic neurons (MZ) in HH10 and HH24 chicken embryo spinal cord. (E)HH10–12 embryos were electroporated with shControl or shRNA for cPHF2 (shcPHF2) cloned into pSUPER vector and GFP-expressing vector. Transversal sections of electroporated neural tubes are indicated above stainedwithDAPI48h postelectroporation (PE). Graphs show the quantification of the size of the control side and shcPHF2-electroporated side. To do so, we measured the dorsal, medial, and ventral distances to the lumen on each side, relative to the length of the central line of the lumen. Data represent the mean of 4 to 5 embryos (from 2 to 4 biological independent experiments). Error bars indicate SD **P < 0.01; ***P < 0.001 (Student’s t test). (F) Transversal sections of neural tubes from HH10–12 embryos electroporated in ovo with shControl or shcPHF2 and stained for H3S10P (PH3), Sox2, or TUJ1 48 PE. Boxplots are showing the quantification of the corresponding immunostaining. Data represent the mean of 4 to 12 embryos (from 3 to 4 biological independent experiments). *P < 0.05 (Student’s t test).

and Pax6 for dorsal progenitors. The results in SI Appendix,Fig. (16–18). To deeply understand the potential contribution of S5E show that PHF2 depletion impairs the generation of all an- PHF2 to neuronal commitment, we first analyzed the PHF2 alyzed cell types. Overall, these results point to an essential role of expression during neuronal differentiation. Our results in SI cPHF2 promoting early neurogenesis by controlling progenitor Appendix, Fig. S6A indicated that PHF2 levels slightly increased proliferation. during differentiation (SI Appendix, Fig. S6A). Next, we analyzed Although we did not observe any effect on neuroblast differ- the consequences of overexpressing PHF2 during neuronal differ- entiation upon PHF2 depletion, it has been demonstrated that entiation that we measured by the induction of the neuronal marker PHF2 plays an important role in differentiation in other models TUJ1 and the repression of the pluripotency gene NESTIN. Results

Pappa et al. PNAS | September 24, 2019 | vol. 116 | no. 39 | 19469 Downloaded by guest on September 29, 2021 in SI Appendix,Fig.S6B and C indicate that PHF2 overexpression variant H2AX is phosphorylated at the Ser-139, forming γH2AX did not affect neuronal differentiation in vitro. as an early cellular response to the DNA double-strand breaks. As PHF2 depletion led to cell cycle arrest, we finally tested Then, we quantified the γH2AX content as a measure of DNA whether PHF2 overexpression overcame G1/S cellular checkpoint damage in control and PHF2-depleted cells. Upon PHF2 KD, a imposed by growth factors removal or neuronal differentiation in- clear accumulation of γH2Ax was detected (Fig. 5A, panel II). duction. Cell cycle reentry was analyzed by measuring the levels of This increase was not observed in cells transduced with the some cell cycle-related genes by qPCR and cell cycle analysis by flow shControl RNA (Fig. 5A, panel I). Interestingly, the observed cytometry. Results in SI Appendix,Fig.S7A–C suggest that, although DNA damage was rescued by overexpression of the PHF2 WT we observed the induction of some genes, PHF2 overexpression did (Fig. 5A, panel III) but not by overexpression of the catalytic dead not overcome the cell cycle arrest imposed by either growth factor mutant (249H > A) (Fig. 5A, panel IV, and Fig. 5B); indeed, in the removal or neuronal differentiation induction. Altogether these data latter case, an apparent increase in γH2Ax reactivity, accompanied strongly indicate that the major role of PHF2 at early neurogenesis is by a higher decrease of replication-related gene expression (SI to facilitate neural progenitor proliferation. Appendix,Fig.S8), was detected, suggesting an important role of PHF2’s enzymatic activity in preventing damage. R-loop (RNA– PHF2 Depletion Leads to DNA Damage and Genome Instability. As DNA hybrids formed by reannealing of nascent transcripts to PHF2 depletion caused profound defects on replication ma- their DNA template, leaving the nontemplated strand as single- chinery expression (E2f, Cyclin E, Orc1/2, and Mcm2/7) as well as stranded DNA) accumulation is associated with DNA replication repair components (Brac1/2, Rad51b, and ATR/M) that finally mistiming that increases collisions with the transcriptional ma- led to cell cycle arrest, we tested whether this fact induced DNA chinery (31), leading to DNA damage (31, 32). Therefore, we in- damage that we measured by the γH2Ax content. The histone vestigated whether γH2Ax reactivity observed in PHF2-depleted

Fig. 5. PHF2 depletion leads to DNA damage and genome instability. (A) Immunostaining of NSCs expressing shControl (I), shPHF2 (II), shPHF2 together with PHF2 WT (III), and shPHF2 together with PHF2 (249H > A) (IV). Cells were fixed and stained with PHF2 and γH2Ax antibodies and DAPI. Enlarged images are showing individual cells. More than 30 cells were quantified. Data shown are representative of 2 to 3 biological independent experiments (Scale bar: 20 μm.) Boxplots represent the number of γH2Ax foci/cell. *P < 0.05; ***P < 0.001 (Student’s t test). (B) Immunostaining of a representative enlarged cell expressing shControl, shPHF2, and shPHF2 together with PHF2 (249H > A). Cells were fixed and stained with γH2Ax and S9.6 antibodies and DAPI. More than 30 cells were quantified. Data shown are representative of 2 to 3 biological independent experiments (Scale bar: 5 μm.) Boxplots represent the number of S9.6 foci/cell. ***P < 0.001 (Student’s t test). (C) Formation of multinucleated cells and segregation defects are shown for shControl and shPHF2 cells. A total of 50 to 100 cells were quantified. Data shown are representative of 4 independent experiments (Scale bar: 5 μm.) Boxplots represent the percentage of multinuclear cells. ***P < 0.001 (Student’s t test).

19470 | www.pnas.org/cgi/doi/10.1073/pnas.1903188116 Pappa et al. Downloaded by guest on September 29, 2021 cells was related to R-loop accumulation. For this purpose, we mice showed partial neonatal death, growth retardation, and performed immunofluorescence experiments in control, PHF2 reduced body weight. Interestingly, the brain weights of Phf2 KD depleted, and PHF2-depleted that overexpressed the PHF2 mice were larger than wild-type littermates (46). Macrocephaly is (249H > A) mutant NSCs (where the γH2Ax signal was higher, see a common phenotype associated with ASD. Intriguingly, we Fig. 5A, panel IV) using the S9.6 antibody that specifically rec- observed that Phf2 KD in cortical progenitors and in the neural ognizes RNA–DNA hybrids without cross-reacting with single-or tube results in defective neural progenitor proliferation. A sim- double-stranded DNA. The data in Fig. 5B show a clear increase ilar paradox has been described for another chromatin factor in S9.6 reactivity that colocalized with the γH2Ax-positive regions. strongly associated with ASD, the remodeling factor CHD8 (47– DNA damage is often linked with genome instability; accordingly, 49). Although at this moment we do not have an explanation for we noticed an increased frequency of segregation de- this phenomenon, it would be interesting to elucidate the role of fects, in particular, anaphase chromatin bridges and multinuclear PHF2 in the development of other cell types such as astrocytes. C cells in PHF2-depleted compared with control NSCs (Fig. 5 ). Furthermore, our results raise the possibility that some of the ’ Altogether these data demonstrate that PHF2 s deficiency induced ASD-associated phenotypes in patients carrying PHF2 mutations DNA damage and R-loop accumulation that led to chromosome may be caused by defects during early neural development. segregation defects and ultimately genome instability. Our work paves the way for investigating the contribution of Discusssion PHF2 to genomic stability and transcriptional regulation in other cellular contexts. In particular, PHF2 has been widely involved in Our data reveal an unforeseen role of PHF2 during develop- cancer. Several studies use H3K9me modifier enzymes as targets ment. We demonstrate that this HDM binds to the cell cycle in cancer treatment (50, 51). Moreover, alteration of the de- gene promoters facilitating their transcription and preventing methylase activity of PHF2 has been suggested as a new target to genome instability. In that way, PHF2 allows neural stem cell treat disorders linked to diet-induced obesity, due to its essential proliferation during progenitor expansion. To do that, PHF2 role in regulating adipogenesis (46). Our results from PHF2- required its catalytic activity. PHF2 prevents H3K9me3 accu- mulation, limiting ectopic heterochromatin formation particu- depleted cells, as well as published data, removing H3K9me2/3 larly at promoter regions. histone methyltransferases (HMTs) (52), indicate that alteration Our data indicate that the histone demethylase PHF2 plays a or removal of H3K9 methylation might not be a suitable thera- pivotal role in the neural development by promoting the pro- peutic strategy. In both cases, genomic instability might be a

liferation of neural progenitors. Interestingly, a synergy for reg- drawback in these treatments. Therefore, our work helps to BIOCHEMISTRY ulation of the key proliferation factor E2F was observed. improve our understanding of the multiple cross-talks between PHF2 directly promotes the expression of several members of epigenetics, development, and diseases. the E2F family and at the same time might cooperate with them Materials and Methods to facilitate their transcriptional activity. These data are in agreement with previous studies reporting that E2F target genes Cell Culture and Differentiation. Mouse NSCs were prepared from cerebral were repressed by H3K9me2/3 marks and HP1 factor (33, 34). cortices of C57BL/6J mouse fetal brains (embryonic day [E]12.5). They were cultured in poly-D-lysine (5 μg/mL, 2 h, 37 °C) and laminine (5 μg/mL, 4 h, Thus, PHF2 might prevent H3K9me2/3 accumulation at the cell 37 °C) precoated dishes (53, 54). Cells were maintained in culture as pre- cycle gene promoters before S entrance, and at the same time it viously described (55) with fibroblast growth factor (FGF) and epidermal facilitates their demethylation to promote cell cycle progression. growth factor (EGF) to 10 and 20 ng/mL, respectively. Human 293T cells were Interestingly, previous studies have shown that the absence of maintained in culture under standard conditions (56). Chicken UMNSAH/ other member of the KDM7 family, PHF8, also impaired G1/S DF1 and HeLa cells were cultured in DMEM supplemented with 10% FBS. transition in conjunction with E2F family factors (29). NSC differentiation protocol is described in SI Appendix, Materials and Our results indicate that PHF2 depletion was associated with Methods. the accumulation of RNA:DNA hybrids, DNA damage, and genome instability. The link between R-loops and DNA damage Antibodies and Reagents. Antibodies used were anti: PHF2 (Cell Signaling, is well established (32). One of the major causes responsible for D45A2), H3K9me3 (Abcam, ab8898), H3K9me2 (Abcam, ab1220), H3K4me3 R-loop accumulation is the replication fork collapse or stall (31, (Abcam, ab8580), H3Sp10 (Millipore, 06-570), anti-phospho-H2A.X (Millipore, β 35–37). Thus, the replication temporal changes due to replica- 07-164), -tubulin III (TUJ1, Covance, MMS-435P), SOX2 (Millipore, AB5603), β α tion machinery alteration generated by PHF2 depletion might -tubulin (Millipore, MAB3408), MNR2 (DSHB, 81.5C10), HP1 (Euromedex), lead to unscheduled collisions of the replication and transcrip- NESTIN (Abcam, ab5968), p-MCM2 (Abcam, ab133243), ATR (Santa Cruz, sc- – 21848), Rad51 (Santa Cruz, sc8349), histones H4 (Abcam, ab10158), H3 tion machineries, inducing DNA breaks (37 40). On the other (Abcam, ab1791), H2B (Abcam, ab1790) and PAX6 (DSHB). Mouse mono- hand, the observed decrease of expression of factors involved in clonal S9.6 is described in ref. 57. DAPI was obtained from Thermo Fisher R-loop resolution and DNA damage repair (ATM/ATR, Rad51, (1306). Doxycyclin (Millipore, 324385) was used at 1 μg/mL. and BRCA1) in PHF2-depleted cells that could contribute to accumulation of double-strand breaks and genome instability. Plasmids. PHF2 cDNA from p3xFLAG-PHF2 (kindly provided by Jiemin Wong, Finally, it has been proposed the demethylation of H3K9 may be East China Normal University, Shangai, China) was cloned into pInducer an important step in the repair of double-strand breaks. Thus, between attB1 and attB2 sites. Both, PHF2 WT and mutant were induced the global increase in H3K9me3 observed in PHF2-depleted upon doxycycline addition (1 μg/mL). pLKO.1 lentiviral vectors expressing progenitors (Fig. 2F) could delay or impair DNA repair, as it short hairpin RNA against mPHF2 (CGTGGCTATTAAAGTGTTCTA), shPHF2 or has been described for KDM4B demethylase (41–43). Interest- control (CAACAAGATGAAGAGCACC) were purchased from Sigma and ingly, PHF2 associated with p53 and facilitates its activity (44). (CCTTATCCACTCCCACTTGACC) shPHF2_2 was cloned in pLKO.1 lentiviral p53 is an essential tumor suppressor that maintains genomic vector. DNA sequences coding cPHF2 short hairpin RNAs (GGAGCTTC- stability. p53, in addition to the cell cycle checkpoint control, GAAGTCGCACT) shcPHF2 and (CTATGTCGGACCAGAGAGA) shcPHF2_2 were cloned into pSUPER or pSHIN vectors (58), as indicated. pSHIN vector con- maintains genomic integrity and replication fidelity by preventing tains the pSUPER and the EGFP expression cassette. DNA topological conflicts between transcription and replication (45). Thus, PHF2 depletion might decrease p53 activity, leading – RNA Extraction and qPCR Assays. RNA was extracted using TRIZOL reagent to accumulation of transcription replication conflicts and genomic (Invitrogen). High-capacity cDNA reverse transcription kit (Invitrogen) and 50 ηg instability. of RNA were utilized for reverse transcription. qPCR assays were performed PHF2 has been related to ASD (22–24), but the function of with SYBR Green (Roche) in a LightCycler 480 (Roche) machine using specific PHF2 in neural development is still poorly understood. Phf2 KD primer pairs (SI Appendix,TableS1).

Pappa et al. PNAS | September 24, 2019 | vol. 116 | no. 39 | 19471 Downloaded by guest on September 29, 2021 Chick in Ovo Electroporation and Immunostaining. Eggs from White-Leghorn samples. Libraries were prepared using the TruSeq Stranded Total RNA chickens were used in the in ovo electroporation experiments. They were Sample Preparation kit with Ribo-Zero Human/Mouse/Rat Kit (Illumina, incubated at 38.5 °C and 70% humidity. Embryo developmental stage was RS-122–2201/2) according to the manufacturer’s protocol. Briefly, 500 ηgof determined following HH (59). Embryos were electroporated with the indicated total RNA was used for ribosomal RNA depletion. Then, after removing ri- DNAs at 3 μg/μLwith50ηg/mL of Fast Green as previously described (54, 60). bosomal RNA, the remaining RNA was fragmented for 4.5 min. The Expanded protocol is described in SI Appendix, Materials and Methods. remaining steps of the library preparation were followed according to the manufacturer’s instructions. Final libraries were analyzed using an Agilent Lentiviral Transduction. Lentiviral transduction was carried out as previously DNA 1000 chip to estimate the quantity and check size distribution, and then described (30, 61). Extended protocol is provided in SI Appendix, Materials quantified by qPCR using the KAPA Library Quantification kit (Roche, and Methods. 07960204001) prior to amplification with Illumina’s cBot. The libraries were sequenced on Illumina High HiSeq 2500 with paired-end 50 long Indirect Immunofluorescence. Immunofluorescence assays were performed reads (SI Appendix, Table S2). Alignment was performed using HISAT2 (64), basically as previously described (61, 62). Cells were fixed in 4% para- assignment of aligned reads to genes was performed using HTSeq (65), and formaldehyde for 20 min and permeabilized using PBS-Triton X-100 (0.1%). differential expression analysis was performed using DESeq2 (66). RNA-seq Then, they were blocked at room temperature for 1 h in 1% BSA (in PBS with data have been deposited in the GEO database under the accession 0.1% Triton X-100). Primary antibodies were incubated overnight at 4 °C. GSE122264 (access token anmrkicwntqjjsp). Finally, Alexa-conjugated secondary IgG antibodies and DAPI (0.1 ηg/μL) (Sigma) were used for 2 h at room temperature. Images were obtained using ChIP-Seq Data Acquisition and Analysis. ChIP-seq data were downloaded from a Leica SP5 confocal microscope by LAS-AF software. Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) (Accessions used in this paper are specified in SI Appendix, Table S3). See further ex- Western Blot. Immunoblotting was performed using standard procedures and planations in SI Appendix, Materials and Methods. visualized using the ECL kit (Amersham). Statistical Analysis. Quantitative data were expressed as mean and SD (for ChIP Assays. ChIP assays were performed as previously described (63). See immunofluoresence countings and RNA transcription experiments) and as extended protocol in SI Appendix, Materials and Methods. mean and SEM (for ChIPs). The significance of differences between groups was assessed using the Student’s t test. Cell Cycle Analysis by Flow Cytometry. Cells were fixed in cold 70% ethanol and stored at −20 °C until they were ready for staining. Cells were washed Data and Materials Availability. RNA-seq and ChIP-seq data have been de- twice with cold PBS and pelleted by centrifugation. The cell pellet was posited in the GEO database under accessions GSE122264 and GSE122263, resuspended in PBS containing 200 μL of 1 mg/mL propidium iodide and respectively. All other materials are available upon request. All the experiments 2 mg RNase A and incubated at 37 °C for 15 min. have been approved by the Bioethics Committee of the Consejo Superior de Investigaciones Científicas. ChIP-Seq Procedure. Chromatin immunoprecipitation as well as sample preparation for sequencing from one replicate were done as previously ACKNOWLEDGMENTS. We thank Drs. Jiemin Wong and Jiwen Li for the described (54). For details see expanded protocol in SI Appendix, Materials PHF2 plasmid; Dr. Borisy for the pSHIN vector; and Drs. A. Jordá, F. Azorín, and Methods and Table S2. ChIP-seq data have been deposited in the GEO J. Bernues, L. L. Espinás, S. de la Luna, A. Vaquero, N. Agell, J. Roig, G. Vicent, M. Beato, M. Arbonés, and E. Martí for reagents; and all members of the database under the accession GSE122263 (access token ovizqewwblullyx). M.A.M.-B. laboratory for helpful comments and suggestions. This study was supported by grants BFU2012-34261 and BFU2015-69248-P (to M.A.M.-B.) RNA-Seq Procedure. RNA was extracted using the High Pure RNA isolation kit from the Spanish Ministry of Economy and Fondation Jérôme Lejeune from Roche followed by DnaseI treatment from 2 biological independent (to M.A.M.-B.).

1. L. Tiberi, P. Vanderhaeghen, J. van den Ameele, Cortical neurogenesis and morpho- 18. J. Yang et al., Epigenetic regulation of megakaryocytic and erythroid differentiation gens: Diversity of cues, sources and functions. Curr. Opin. Cell Biol. 24, 269–276 (2012). by PHF2 histone demethylase. J. Cell. Physiol. 233, 6841–6852 (2018). 2. S. Temple, The development of neural stem cells. Nature 414, 112–117 (2001). 19. D. R. Pattabiraman et al., Activation of PKA leads to mesenchymal-to-epithelial 3. T. Kouzarides, Chromatin modifications and their function. Cell 128, 693–705 (2007). transition and loss of tumor-initiating ability. Science 351, aad3680 (2016). 4. R. Fueyo, M. A. García, M. A. Martínez-Balbás, Jumonji family histone demethylases in 20. A. Ghosh et al., Association of FANCC and PTCH1 with the development of early neural development. Cell Tissue Res. 359,87–98 (2015). dysplastic lesions of the head and neck. Ann. Surg. Oncol. 19 (suppl. 3), S528–S538 5. S. Epsztejn-Litman et al., De novo DNA methylation promoted by G9a prevents re- (2012). – programming of embryonically silenced genes. Nat. Struct. Mol. Biol. 15, 1176 1183 21. C. Lee, B. Kim, B. Song, K. C. Moon, Implication of PHF2 expression in clear cell renal (2008). cell carcinoma. J. Pathol. Transl. Med. 51, 359–364 (2017). 6. B. Wen, H. Wu, Y. Shinkai, R. A. Irizarry, A. P. Feinberg, Large histone H3 lysine 9 di- 22. R. K. C Yuen et al., Whole genome sequencing resource identifies 18 new candidate methylated chromatin blocks distinguish differentiated from embryonic stem cells. genes for autism spectrum disorder. Nat. Neurosci. 20, 602–611 (2017). – Nat. Genet. 41, 246 250 (2009). 23. I. Iossifov et al., De novo gene disruptions in children on the autistic spectrum. Neuron 7. G. J. Filion, B. van Steensel, Reassessing the abundance of H3K9me2 chromatin do- 74, 285–299 (2012). mains in embryonic stem cells. Nat. Genet. 42, 4; author reply 5-6 (2010). 24. J. Cotney et al., The autism-associated chromatin modifier CHD8 regulates other 8. J. Nakayama, J. C. Rice, B. D. Strahl, C. D. Allis, S. I. Grewal, Role of histone H3 lysine autism risk genes during human neurodevelopment. Nat. Commun. 6, 6404 (2015). 9 methylation in epigenetic control of heterochromatin assembly. Science 292, 110– 25. J. Bricambert et al., The histone demethylase Phf2 acts as a molecular checkpoint to 113 (2001). prevent NAFLD progression during obesity. Nat. Commun. 9, 2092 (2018). 9. R. C. Allshire, J. P. Javerzat, N. J. Redhead, G. Cranston, Position effect variegation at 26. X. N. Wu et al., Methylation of transcription factor YY2 regulates its transcriptional fission yeast centromeres. Cell 76, 157–169 (1994). activity and cell proliferation. Cell Discov. 3, 17035 (2017). 10. N. Saksouk, E. Simboeck, J. Déjardin, Constitutive heterochromatin formation and 27. T. Lammens, J. Li, G. Leone, L. De Veylder, Atypical E2Fs: New players in the E2F transcription in mammals. Epigenetics Chromatin 8, 3 (2015). transcription factor family. Trends Cell Biol. 19, 111–118 (2009). 11. Y. Okuno, K. Inoue, Y. Imai, Novel insights into histone modifiers in adipogenesis. 28. S. Tahmasebi et al., Control of embryonic stem cell self-renewal and differentiation Adipocyte 2, 285–288 (2013). 12. K. Fortschegger, R. Shiekhattar, Plant homeodomain fingers form a helping hand for via coordinated alternative splicing and translation of YY2. Proc. Natl. Acad. Sci. – transcription. Epigenetics 6,4–8 (2011). U.S.A. 113, 12360 12367 (2016). 13. H. Wen et al., Recognition of histone H3K4 trimethylation by the plant homeodomain 29. W. Liu et al., PHF8 mediates histone H4 lysine 20 demethylation events involved in cell – of PHF2 modulates histone demethylation. J. Biol. Chem. 285, 9322–9326 (2010). cycle progression. Nature 466, 508 512 (2010). 14. G. A. Nicholson et al., The gene for hereditary sensory neuropathy type I (HSN-I) maps 30. E. Asensio-Juan et al., The histone demethylase PHF8 is a molecular safeguard of the to chromosome 9q22.1-q22.3. Nat. Genet. 13, 101–104 (1996). IFNγ response. Nucleic Acids Res. 45, 3800–3811 (2017). 15. K. Hasenpusch-Theil et al., PHF2, a novel PHD finger gene located on human chro- 31. W. Gan et al., R-loop-mediated genomic instability is caused by impairment of rep- mosome 9q22. Mamm. Genome 10, 294–298 (1999). lication fork progression. Genes Dev. 25, 2041–2056 (2011). 16. H. J. Kim et al., Plant homeodomain finger protein 2 promotes bone formation by 32. J. Sollier, K. A. Cimprich, Breaking bad: R-Loops and genome integrity. Trends Cell demethylating and activating Runx2 for osteoblast differentiation. Cell Res. 24, 1231– Biol. 25, 514–522 (2015). 1249 (2014). 33. S. J. Nielsen et al., Rb targets histone H3 methylation and HP1 to promoters. Nature 17. K. H. Lee, U. I. Ju, J. Y. Song, Y. S. Chun, The histone demethylase PHF2 promotes fat 412, 561–565 (2001). cell differentiation as an epigenetic activator of both C/EBPα and C/EBPδ. Mol. Cells 34. I. Panteleeva et al., HP1alpha guides neuronal fate by timing E2F-targeted genes si- 37, 734–741 (2014). lencing during terminal differentiation. EMBO J. 26, 3616–3628 (2007).

19472 | www.pnas.org/cgi/doi/10.1073/pnas.1903188116 Pappa et al. Downloaded by guest on September 29, 2021 35. S. Tuduri et al., Topoisomerase I suppresses genomic instability by preventing in- 50. Y. Kondo et al., Downregulation of histone H3 lysine 9 methyltransferase G9a induces terference between replication and transcription. Nat. Cell Biol. 11, 1315–1324 (2009). centrosome disruption and chromosome instability in cancer cells. PLoS One 3, e2037 36. A. Helmrich, M. Ballarino, L. Tora, Collisions between replication and transcription (2008). complexes cause common fragile site instability at the longest human genes. Mol. Cell 51. T. Wagner, M. Jung, New lysine methyltransferase drug targets in cancer. Nat. Bio- 44, 966–977 (2011). technol. 30, 622–623 (2012). 37. A. M. Deshpande, C. S. Newlon, DNA replication fork pause sites dependent on 52. P. Zeller et al., Histone H3K9 methylation is dispensable for Caenorhabditis elegans transcription. Science 272, 1030–1033 (1996). development but suppresses RNA:DNA hybrid-associated repeat instability. Nat. 38. E. A. Hoffman, A. McCulley, B. Haarer, R. Arnak, W. Feng, Break-seq reveals Genet. 48, 1385–1395 (2016). hydroxyurea-induced chromosome fragility as a result of unscheduled conflict be- 53. D. S. Currle, J. S. Hu, A. Kolski-Andreaco, E. S. Monuki, Culture of mouse neural stem – tween DNA replication and transcription. Genome Res. 25, 402 412 (2015). cell precursors. J. Vis. Exp., e152 (2007). 39. H. Merrikh, C. Machón, W. H. Grainger, A. D. Grossman, P. Soultanas, Co-directional 54. R. Fueyo et al., Lineage specific transcription factors and epigenetic regulators me- – replication-transcription conflicts lead to replication restart. Nature 470, 554 557 diate TGFβ-dependent enhancer activation. Nucleic Acids Res. 46, 3351–3365 (2018). (2011). 55. C. Estarás et al., Genome-wide analysis reveals that Smad3 and JMJD3 HDM co- 40. A. Bayona-Feliu, A. Casas-Lamesa, O. Reina, J. Bernués, F. Azorín, Linker histone activate the neural developmental program. Development 139, 2681–2691 (2012). H1 prevents R-loop accumulation and genome instability in heterochromatin. Nat. 56. N. Blanco-García, E. Asensio-Juan, X. de la Cruz, M. A. Martínez-Balbás, Autoacety- Commun. 8, 283 (2017). lation regulates P/CAF nuclear localization. J. Biol. Chem. 284, 1343–1352 (2009). 41. S. U. Colmenares et al., Drosophila histone demethylase KDM4A has enzymatic and 57. S. J. Boguslawski et al., Characterization of monoclonal antibody to DNA.RNA and its non-enzymatic roles in controlling heterochromatin integrity. Dev. Cell 42, 156– application to immunodetection of hybrids. J. Immunol. Methods 89, 123–130 (1986). 169.e5 (2017). 58. S. Kojima, D. Vignjevic, G. G. Borisy, Improved silencing vector co-expressing GFP and 42. L. C. Young, D. W. McDonald, M. J. Hendzel, Kdm4b histone demethylase is a DNA small hairpin RNA. Biotechniques 36,74–79 (2004). damage response protein and confers a survival advantage following γ-irradiation. J. 59. V. Hamburger, H. L. Hamilton, A series of normal stages in the development of the Biol. Chem. 288, 21376–21388 (2013). chick embryo. 1951. Dev. Dyn. 195, 231–272 (1992). 43. H. Zheng, L. Chen, W. J. Pledger, J. Fang, J. Chen, p53 promotes repair of hetero- 60. N. Akizu et al., EZH2 regulates neuroepithelium structure and neuroblast pro- chromatin DNA by regulating JMJD2b and SUV39H1 expression. Oncogene 33, 734– 744 (2014). liferation by repressing p21. Open Biol. 6, 150227 (2016). 44. K. H. Lee et al., PHF2 histone demethylase acts as a tumor suppressor in association 61. E. Asensio-Juan, C. Gallego, M. A. Martínez-Balbás, The histone demethylase PHF8 is – with p53 in cancer. Oncogene 34, 2897–2909 (2015). essential for cytoskeleton dynamics. Nucleic Acids Res. 40, 9429 9440 (2012). 45. C. Q. X. Yeo et al., p53 maintains genomic stability by preventing interference be- 62. S. Sánchez-Molina et al., Regulation of CBP and Tip60 coordinates histone acetylation tween transcription and replication. Cell Rep. 15, 132–146 (2016). at local and global levels during Ras-induced transformation. Carcinogenesis 35, 46. Y. Okuno et al., Epigenetic regulation of adipogenesis by PHF2 histone demethylase. 2194–2202 (2014). Diabetes 62, 1426–1434 (2013). 63. E. Valls et al., Involvement of chromatin and histone deacetylation in SV40 T antigen 47. B. J. O’Roak et al., Multiplex targeted sequencing identifies recurrently mutated transcription regulation. Nucleic Acids Res. 35, 1958–1968 (2007). genes in autism spectrum disorders. Science 338, 1619–1622 (2012). 64. D. Kim, B. Langmead, S. L. Salzberg, HISAT: A fast spliced aligner with low memory 48. S. De Rubeis et al.; DDD Study; Homozygosity Mapping Collaborative for Autism; requirements. Nat. Methods 12, 357–360 (2015).

UK10K Consortium, Synaptic, transcriptional and chromatin genes disrupted in au- 65. S. Anders, P. T. Pyl, W. Huber, HTSeq–A Python framework to work with high- BIOCHEMISTRY tism. Nature 515, 209–215 (2014). throughput sequencing data. Bioinformatics 31, 166–169 (2015). 49. R. Bernier et al., Disruptive CHD8 mutations define a subtype of autism early in de- 66. M. I. Love, W. Huber, S. Anders, Moderated estimation of fold change and dispersion velopment. Cell 158, 263–276 (2014). for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).

Pappa et al. PNAS | September 24, 2019 | vol. 116 | no. 39 | 19473 Downloaded by guest on September 29, 2021